1. Introduction

Fiber optic sensors (FOS) offers many advantages over conventional sensors, such as corrosion resistance, multiplexing, compact design, and immunity from many types of electromagnetic radiations [1–4]. Because of these advantages, FOS is a very active research field covering a host of applications: biomedical applications, vibrational detection, bridge monitoring, and temperature sensing [4–8]. In the case of temperature sensing, many types of FOS have been investigated; however, most of these sensors suffer from some limitations.

Fiber Bragg grating (FBG) sensors are a popular commercially available FOS. FBGs are compact, reliable, and easy to integrate into existing systems. However, FBG require expensive interrogation equipment and standard Type I gratings have a maximum temperature of only a few hundred degrees Celsius [9,10], limiting the possible applications that FBG can be used for. For higher temperatures, photonic crystal fibers (PCF) have received considerable attention for their ability to operate stably up to 1100 °C [11,12]. These high temperature sensors are created by splicing a section of PCF between standard single mode fiber (SMF), resulting in a Fabry-Perot type interferometer [13,14]. However, these types of PCF interferometers require complex fabrication including offset splicing and collapsing of sections of the PCF splice [11,13]. Recently, multicore fiber (MCF) based FOS have been fabricated that are thermally-stable at high temperature (1000 °C), cost efficient, and easy to fabricate; offering an attractive alternative to FGB and PCF [15–17].

Amezcua-Correa et al. [15–17] demonstrated that a section of MCF spliced between two SMF sections, similar to the PCF-type sensors, can be used to accurately measure temperature. The MCF is comprised of seven cores arranged in a symmetric hexagonal shape, Fig. 1

 

Fig. 1 Seven Core MCF (a) schematic of MCF with hexagonal core shape, (b) microscope image of MCF, (c) microscope image with small reflection on bottom right that clearly shows the geometry of the MCF, and (d) periodic interference pattern of SMF-MCF-SMF device.

Download Full Size | PDF

The purpose of this study was to produce a commercially viable MCF temperature sensor operating in reflection mode. An MCF sensor was packaged into a 0.559 mm diameter stainless steel tube, Fig. 2

 

Fig. 2 Experimental setup consists of a light source, fiber optical circulator, packaged MCF temperature sensor, and spectrometer or optical interrogator (depending on experiment). MCF temperature sensor consists of a 5mm section of MCF fusion spliced between two SMFs. The Sensor is packaged into a stainless tube and operated in reflection mode.

Download Full Size | PDF

2. Sensor fabrication

The MCF temperature sensor was fabricated by fusion splicing a 5 mm section of MCF between two sections of SMF, Fig. 2. The MCF was 3rd generation MCF, purchased from The College of Optics and Photonics (CREOL) at University of Central Florida. To operate in reflective mode, the end of one of the SMFs was coated with a reflective metal. The fiber was then packaged into a stainless tube of 0.559 mm diameter. The stainless-steel tubing serves two purposes; it protects the bare-fiber from physical damage and it provides a support, preventing the fiber from moving. Preventing movement of the fiber is important because the same SMF-MCF-SMF sensor can also be used to measure force, strain, and curvature [16,18,19]. The stainless-steel tube prevents the fiber from moving and causing non-temperature related changes in the interference spectra. The bare fiber was not constrained to the inside of the tube, there is a slight clearance between the tube and fiber. Previous efforts to pot the fiber using ceramic cements resulted in noisy, non-usable data. We believe this is because of mis-matching thermal expansion between the fiber and cement causing stain and/or bending on the sensor. The non-coated SMF is terminated to a standard FC connector, Fig. 2.

3. Temperature detection

3.1 Spectrometer calibration

To determine the quality of the packaged MCF temperature sensor, the temperature response to the multimode interference was investigated. The experimental setup is shown in Fig. 2. The sensor was connected to a stabilized tungsten IR light source (Thorlabs, SLS202L), a fiber optic circulator, and a spectrometer (Ocean Optics, NIRQuest512-2.2). The sensor was placed into a stream of heated air of varying temperature, and the change to the spectra was observed. Figure 3(a)

 

Fig. 3 (A) Spectral shift of MCF. With increasing temperature, the wavelength of maximum reflection shifts to larger wavelengths. (B) Linear plot shows a spectral shift 35 pm/°C.

Download Full Size | PDF

3.2 Photodiode calibration

Although the spectrometer worked well in demonstrating the spectral response to temperature changes in the MCF, a cheaper alternative is required if MCF sensors are to be widely distributed commercially. To this end, a low-cost optical integrator was fabricated from commercially available off-the-shelf parts, with a photodiode used as a detector. The photodiode converts light into an electrical current, and the current can be used to calibrate the packaged MCF sensor. To test the effectiveness of the photodiode-based interrogator, the packaged MCF sensor was testing in both a liquid and gas medium.

Experiments involving the photodiode-based interrogator used a monochromatic light source of 1525 nm, produce from a Golight Tunable Light Source (Shenzhen Golight Technology Co.). A mass-produced interrogator can be little more than a fiber coupled telecom laser. The choice of wavelength is not arbitrary and should be chosen to fall between a peak or valley of the interference spectrum. This is because the calibration is based on the photodiode, which outputs a value that is proportional to the reflection value, i.e. the larger the reflection percent, the larger the output of the photodiode. If the wavelength of the light source is chosen such that it falls on a peak, the reflection value will decrease with both heating and cooling. Conversely, if the wavelength of the light source is chosen such that it falls on a valley, the reflection value will increase with both heating and cooling. For example, at 200 °C the interference pattern has a λmax of 1486 nm. If 1485 nm is chosen as the wavelength of the light source, heating would cause the peak to shift 35 pm/°C [which is the slope in Fig. 3(b)] to the right, and cooling would cause the peak to shift 35 pm/°C left. A temperature of 225 °C (200 + 25) would give the same reflection value as 175 °C (200 - 25). And identical reflection values would give identical outputs from the photodiode detector. For this reason, the wavelength of the light source must be a reasonable distance from a peak or valley value in the interference spectrum.

The photodiode-based interrogator was tested by heating and cooling the MCF temperature sensor in a gas and liquid media, a hot air steam and wave oil, respectively. Figure 4

 

Fig. 4 Temperature response of MCF temperature sensor using photodiode interrogator in a stream of heated air. (A) Counts from photodiode output on right axis, temperature (°C) on left axis. (B) The oscillations in the counts (right axis) is a product of the oscillation of the power supplied to the heating element. The PID loop of the heating element is constantly changing power (left axis), and the MCF response time is fast enough to detect these slight changes in temperature.

Download Full Size | PDF

 

Fig. 5 (A) MCF temperature calibration using photodiode interrogator in a gas (▲) and liquid (■) environment. (B) Error analysis demonstrated the calibration was within 5% error for all points. The line is not fit to data, it is the y = x line, and used for visual aid of calibration precision. The closer a data point is to the line, the more precise the calibration.

Download Full Size | PDF

Additionally, the error was shown to be <5% of the value for all points. Figure 5(b) shows the predicted temperatures plotted against the actual temperatures with a y = x line. The y = x line represents a calibration with zero error, the farther a data point is from the line, the greater the error. This demonstrates the practicality of using an inexpensive photodiode-based interrogator with an inexpensive MCF temperature sensor.

The differences in the calibration equations are because the sensor was calibrated at different reference temperatures on different days. Many photodiode characteristics, including the output current, are temperature dependent. The temperature in our lab is not constant, and the calibrations in Fig. 5 were performed on different days. These difference in temperatures account for the differences in calibration equations for the liquid and gas experiments. Future work involves software integration that accounts for reference temperature of the photodiode interrogator; analogous to the cold junction temperature adjustments of thermocouples [20].

One important metric in temperature sensing is response time. A common measurement to compare response times between sensors is the Time Constant (τ_63%), which is the time required to reach 63.2% of an instantaneous change in temperature. Figure 6

 

Fig. 6 Response time of the packaged MCF temperature sensor and the theoretical response time of a k-type thermocouple of identical dimensions. The MCF sensor has a τ_63% of 0.09 seconds, while the TC has a theoretical τ_63% > 1.0 seconds. The TC time constant was calculated from correlation charts provided by Omega [21]. The line for the TC is not meant to represent a liner relationship, the data is a starting value and a final value. The line was added for visual aid only. The TC time constant was calculated from correlation charts provided by Omega [21]. The line for the TC is not meant to represent a liner relationship, the data is a starting value and a final value. The line was added for visual aid only.

Download Full Size | PDF

4. Conclusion

In conclusion, an MCF-based temperature sensor was fabricated by splicing a 5 mm section of MCF between two sections of SMF. The resulting fiber was then packaged in a 0.559 mm diameter stainless steel tube, which provides a commercial friendly formfactor. The MCF interference pattern was shown to have a linear 35 pm/°C shift to changes in temperature. The package MCF sensor was also tested and calibrated using a photodiode-based interrogator in both liquid and gas mediums, which also showed a linear temperature response. The response time was measured and compared to the theoretical response time of a TC of identical diameter. It was found that the MCF temperature sensor has a time constant of 0.09 seconds, compared to over 1.0 second for a thermocouple. To the best of our knowledge, this is the first time that a packaged MCF temperature device has been reported, all other reports used an exposed bare-fiber sensor. Additionally, this is the first known reporting which used a photodiode-based interrogator to measure the response from the MCF sensor. The combination of these two firsts, demonstrates the feasibility of an inexpensive and highly responsive fiber optic-based temperature sensor. Future work includes creating a reference temperature adjustment for the photodiode, and a complete determination of important characteristics: resolution, sensitivity, and full temperature range.